1j.j ,4s1i X $11els!*lt#tiWt 'irl t l m :!$!t; l 91,,1jtt !* S hg W t SS iW -t B! sitdez A m A!b

1iilIiltIi113SB;|1101111RARIMM|/saIRalIUMINXRSr -- =XE= §S%iiiiiii i ii iii ii;]i iiiR t lsrtAi ; 1 ;;1tiil_9_ibEsS _ lfi 1 1 - _ :-:- =s rR 1m I, lrWUlgl:Xtt;gtsjplSn5s.-- ............. Ew r--W-_ w ......._Z _. =

2 -_ i95 ; D -F j-'nl9 E !i--55! j =31't1 ilBl '' :. j "jjj -i-='i_ S ii ----t ''----- --

_ | R _ -1 | i 1 i i 1 5 i l t . - - _ _ '~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~!- = _ _ _ _

_

i nKwS s; scASur ;ur M i ~I r............................. ....---5- ~-Y

i _ i!l.3il,li!l,!ir! i ii Si li ii 15ii [ -i ;-j_ -ij jj j rl --5 i ~~~~~ i jj 5 j j j jj jjj j jj j jj ;;;;; ---- ---------~= L-- :. -: -:;::;; ;;. ; .c ;; : ; 4 4 ; 4;;;;;;;;;c n,;; - ; r ; ;c. -a =_ __ = _e ____

_ce, w>[|Xp -,,R«iM t ! _lR llI 'e!l'ti!exDI! sl llXl!!! P!!!W4" t il!llA X'e"R Tti- -----------------------------------------

lffitufih -|w":litlllzl11D91|1~~~~~ Jl11D1=.'%lAl!l!l^fil o IWill*iM 15t' e '"e

UNDP/UNCHS/World BankUrban Management Programme

Urban Management and Municipal Finance

16

Decentralization and its Implicationsfor Service Delivery

W lliam Dillinger

Published for the Urban Management Programme byThe World Bank, Washington, D.C.

: .,~~

This document has been prepared under the auspies of the United Nations Development Programme/UnitedNations Center for Human Settlements (Habitat)/World Bank-sponsored Urban Management Programme. Thefindings, interpretations, and conclusionsexpressed hereare those of the authorsand do not necessarily representthe views of the United Nations Development Programme, uNciHS the World Bank, or any of their affiliatedorganizations.

Deputy Director Chief ChiefDivision for Global Technical Co-operation Urban Development Division

and Interregional Programmes Division Transport, Water, and UrbanUnited Nations Development United Nations Center Development Department

Programme for Human Settlements Environmentally Sustainable(Habitat) Development

The World Bank

Copyright © 1994The Intemational Bank for Reconstructionand Development/HE WORLD DANK1818 H Street, N.W.Washington, D.C. 20433, USA

All rights reservedManufachured in the United States of AmericaFirst printing May 1994

The Urban Management Programme (uw) represents a major approachby the United Nations family of organizations,ogedt with external support agencies (ESAs), to strengthen the contnbution that cities and towns in developing

countries make toward economic growth, social development, and tie alleviation of poverty. The program seeks todevelop and promote appropriate policies and tools for municipal finance and administration, land management,infrastruchtue management, and environmental management. Through a capacity building component, the UMP plansto establish an effective partnership with national, regional, and global networks and ESAs in applied research,dissemination of information, and experiences of best practices and promising options.

The findings, interpretations, and cnclusions expressed in this paper are entirely those of the author(s) and should notbe attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of ExecutiveDiets or the countries tiey represent The World Bank does not guarantee the accuracy of the data included in thispublication and accepts no responsibility whatsoever for any consequence of their use. Some sources cited in tiis papermaybe informal documents thatarenotreadily available Theboundaries, colors, denoniinations, and other informationshown on any map in this volume do not imply on the part of the World Bank Group anyjudgment on the legal statusof any territory or the endorsement or acceptance of suth boundaries.

The material in this publication is copyrighted. Requests for permission to reproduce portions of it should be sent to theOffice of the Publisher at the address shown in the copyright notice above. The World Bank encourages disserinationof its work and will normally give permission promptly and, when the reproduction is for noncommercial purposes,without asking a fee. Permission to copy portions for classroom use is granted through the Copyright Clearance Center,Inc., Suite 910, 222 Rosewood Drive, Danvers, Massachusetts 01923, USA

ISSN: 1020-0215

libray of Congress Cataloging-in-Publication Data

Dillinger, William R., 1951-Decentralization and its implications for urban service delivery /

William Dillinger.p. cn. - (Urban management progranmne, ISSN 1020-0215- Urban

management and municipal finance; 16)Indudes bibliographical reerenetISBN 0-8213-2792-51. Municipal services-Developing countries 2. Decentralization

in government-Developing countries. 3. Intergovernmental fiscalrelations-Developing countries. L Title. IL Series.HD4431.D55 1994363'09172'4-dc2O 94-5013

C?

CONTENTS

EXECUTIVE SUMMARY I

L INTRODUCTION 5The Context 5The Management Problem 7

II. A FRAMEWORK FOR ANALYSIS 11Defining the Problem 11Defining the Instruments 13A Typology of Intergovernmental Relations 16

IIL DIRECTIONS FOR REFORM 24Linking Services to Levels of Government 24Linking Revenues to Expenditures 26Balancing Regulation with Political Accountability 33Synchronizing the Elements of Reform 37

REFERENCES 38

BOXEBox 1. A typology of decentralization 7Box 2. Contrasting approaches to capital allocation 23Box 3. Choosing among local benefit taxes 29Box 4. Improving the performace of MCIs 34

TABLESTable 1. Trends in world urbanization 5Table 2. Sectoral composition of economic growth 5Table 3. Structure of political accountability in major LDC cities 18Table 4. Sources of tax revenue in major LDC cities 19Table S. Municipal lending organizations 33

FIGUREFigure 1. Local share of public expenditure 17

FOREWORD

This paper has been prepared for the Municipal Finance component of the joint UNDP/UNCHS/World Bank-Urban Management Progrnmme (UMP). The UMP represents a majorapproach by the United Nations family of organizations, together with external support agencies(ESAs), to strengthen the contribution that cities and towns in developing countries make towardseconomic growth, social development, and the alleviation of poverty. The program seeks to developand promote appropriate policies and tools for municipal finance and administration, land manage-ment, infrastrmcture management, environmental management, and poverty alleviation. Through acapacity building component, the UMP plans to establi.h an effective partnership with national,regional, and global networks and ESAs in applied research, dissemination of infornation, andexperiences of best practices and promising options.

This paper is one in a series of discussion papers and management tools to be producedby the UMP municipal finance component. As a whole the municipal finance component is intendedto address three cuestions: 1) how to mobilize resources to finance the delivery of urban services; 2)how to improve the financial management of those resources; and 3) how to organize municipalinstitutions to promote greater efficiency and responsiveness in urban service delivery. Work duringthe initial phase of the Urban Management Programme has focused on the first of these questions-focusing specifically on local tax reform, intergovernmental transfers, and local access to long-termcrediL Case studies and background papers on the latter questions-documenting issues in localfinancial management and the organization of municipal government-have also been prepared, andwill provide the basis for publications to be issued under this series in the future.

Phase 2 of the UMP (1992-96) is concerned with capacity building at both the countryand regional levels and with facilitating national and municipal dialogues on policy and programoptions. It emphasizes a participatory structure that draws on the strengths of developing countryexperts and expedites the dissemination of that expertise at the local, national, regional, and globallevels.

Through its regional offices in Africa, the Arab States, Asia and the Pacific, and LatinAmerica and the Caribbean, the UMP seeks to strengthen urban management by hamessing the skillsand strategies of regional experts, communities, and organizations in the private sector.

Regional coordinators use these networks to address the five program themes in twoways:

City and country consultations The UMP brings together national and localauthorities, private-sector networks, community representatives, and otheractors to discuss specific problems within the UMP's subject areas and topropose reasoned solutions. Consultations are held at the request of a country orcity, and often provide a forum for discussion of a cross-section of issues.

vi

Technical cooperation. To sustain follow-up to the consultations, the UMP usesits regional networks of expertise to provide technical advice and cooperadon.

Through its nucleus team in Nairobi and Washington, D.C., the UMP supports itsregional program and networks by synthesizing lessons leamed, conducting state-of-the-art research,and supporting dissemination of program related materials.

MarkHildebrand Louis Y. PouliquenChief DirectorTechnical Cooperation Division Transportation, Water, andUnited Nations Centre for UrbanDevelopmentDepartmentHuman Setdements (HABITAT)

ABSTRACT

This paperreviews efforts to improve theefficiency and responsivenessof urban servicedelivery indeveloping countries. Itargues thatfailures in urban servicedelivery are notmerely the resultof a lack of technical knowledge on the part of local government staff, but also reflect constraints andperverse incentives confronting local personnel and theirpolitical leadership, and that these, in turn, areoften the inadvertent result of problems in the relationship between central and local government.

In this respect, the report views the spread of decentralization as a potentially fortuitousphenomenon. As apolitical phenomenon, decentralization is widespread. Out of the 75 developing andtransitional countrieswith populations greater than 5 million, all but 12 claim to be embarked on someform oftransferofpolitical powertolocal units ofgovernment. But theobjectives ofdecentralization-as it is observed in practice-appear only tangentially related to administrative performance. Thedecentralizationnow occurring isnot acarefully designed sequenceofreforms aimed atimproving theefficiency of public service delivery; it appears to be a reluctant and disorderly series of concessionsby central governments attempting to maintain political stability.

Nevertheless, itpresents reformers-both domestic and in the donorcommunity-with anopportunity to promote the kinds of fundamental reforms that have proven frustrating in the past.Because decentralizationhasintroducedahighdegreeoffluidityintothestructureofintergovernmentalrelations, it has brought flexibility into what had appeared to be an immutable system of governance.

The stakes are high. Decentralization affects not only urban services, but also socialsectors, non-urban infrastructure, and-conceivably-the stability of national economiies and theeffectiveness of poverty-alleviation efforts. As the present degree of fluidity in intergovernmentalrelations is presumably transitory, it is an opportunity that should be seized.

ACKNOWLEDGMENTS

In preparing this paper, the author has greatly benefited from the advice of Arturo Israeland Richard Bird and on the assistance of his colleagues within the Urban Development division:Victor Vergara, Rita Hilton, Fitz Ford, andJim Hicks. The paper further benefited from the commentsof the following external reviewers: Ken Davey (The University of Birmingham) and Fernando Rojas(ILSA, Colombia). The work was carried out under the direction of L.Y. Pouliquen, P. Annez, andM. Cohen. All errors and omissions remain the responsibility of the author.

Thispaperbuilds upon two decades of workby World Bankstaffin the fieldofmunicipaldevelopment. It has drawn extensively on the work on urban public finance by Johannes Linn andRoy Bahl, the work on institutional development by Arturo Israel, Samuel Paul, and Elinor Ostromand on the more recent studies on decentralization and intergovemmental fiscal relations by JerrySilverman, Tim Campbell, Don Winkler, and Anna Haines. A major source of information-bothon past practice and on current innovations-has been project officers directly involved in urbanlending operations.

EXECUTIVE SUMMARY

Introduction

i. Thispaperreviewsefforts to improvethe efficiencyand responsiveness of urban servicedelivery in developing countries. It argues that failures in urban service delivery are not merely theresult of a lack of technical knowledge on the part of local government staff, but also reflectconstraints and perverse incentives confronting local personnel and theirpolitical leadership, and thatthese, in tum, are often the inadvertent resultof problems in the relationship between central and localgovernment.

ii. In this respect, the report views the spread of decentralization as a potentially fortuitousphenomenon. As a political phenomenon, decentralization is widespread. Out of the 75 developingand transitional countries with populations greater than 5 million, all but 12 claim to be embarkedon some form of transfer of political power to local units of government. But the objectives ofdecentralization-as it is observed in practice-appear only tangentially related to administrativeperformance. The decentralization now occurring is not a carefully designed sequence of reformsaimed atimproving theefficiency ofpublic service delivery; itappears to be areluctantand disorderlyseries of concessions by centmr governments attempting to maintain political stability.

Mi. Nevertheless, itpresents refonner-both domestic and in the donorcommunity-withan opportunity to promote the kinds of fundamental reforms that have proven frustrating in the past.Because decentralization has introduced a high degree of fluidity into the structure of intergovern-mental relations, it has brought flexibility into what had appeared to be an immutable system ofgovernance.

iv. The stakes are high. Decentralization affects not only urban services, but also socialsectors, non-urban infrastructure, and-conceivably-the stability of national economies and theeffectiveness of poverty-alleviation efforts. As the present degree of fluidity in intergovemmentalrelations is presumably transitory, it is an opportunity that should be seized.

Directions for Reform

v. The basis for mkidng recommendations on the structure of urban service delivery isnecessarily polyglot. The academic literature provides a useful conceptual fiamework. The publiceconomics literature, for example, provides a basic prescription for expenditure and revenueassignment, arguing that responsibility for discrete public services should be assigned to the level ofgovernment whose boundaries incorporate the affected beneficiaries-along with a correspondingrevenue source with which to ascertain demand. The institutional analysis literature contributes amore extensive list of policy handles, a view that the performance of organizations is not onlydetennined by their nominal powers and responsibilities, but by the motivations of the people whowork within them. This directs attention to such issues as whether the mayor is appointed or elected(and what sort of career trajectory confronts a successful mayor), and the mechanisms by whichinterest groups can make their wishes known.

2

vi. The mostpersuasive basis forrecommendationsisempirical evidence, but the empiricalevidence that exists is not decisive. Evidence from developed countries-which might be presumedto provide models-provides some common sense of direction but is characterized by considerablevariety and frequentexperimentation. The past experience from developing countries appears largelyto be a source of negative lessons. The most relevant body of evidence is emerging from the countriesthat are now in the process of decentralization. But these reforns are still in their initial stages. Thereare no before-and-after cases of LDC decentralization on which to base nornative advice.

vii. The evidence nevertheless suggests that there are three elements to reform in thestructure of urban service delivery: (a) the clarification of functional responsibilities between levelsof govemment; (b) the authorization of revenue sources corresponding to functional responsibilities,and (c) the institution of a system of accountability that encompasses both regulation by centralgovernment and incentives for responsiveness to local constituents.

Linking services to government

viii. Clarity in the division of functional responsibilities between levels of governmentwould appear to be an essential condition of any reform in the structure of urban service delivery. Aclear linkage between a unit of government and a specific service seems to be essential to enableconstituents to hold local governments accountable for specific functions and to eliminate the softbudget constraint implied by the otherwise open-ended nature of central government participation.Such clarity requires, interalia, asystem of revenue assignmentand budgetary discretion that permitslocal governments to perform the roles that have been assigned to them. But it may first requiregovemments to legislate more geographic specificity into their municipal organic laws.

Revenue reform

ix. Reform in revenue assignments is needed if a clear division of functional responsibilitiesisto be workable.The particularstructure of local revenuese mix ofusercharges, taxes, transfers,and loans-that is appropriate in a given oontext depends, first and foremost, on the functions thathave been assigned to local governmenL Where the benefits of a service are largely confined toindividual consumers, user charges are an attractive means of financing municipal services. Localtaxes, in principle, are an appropriate means of financing services whose benefits cannot be confinedto individual consumers, but nevertheless do not extend beyond the municipal boundaries.

x. Any attempttoreform the structure of urban servicedelivery,however, mustaddress thelargest source of local revenue: intergovemmental transfers. Transfers can serve several importantpositive roles in the financing of municipal services, permitting central govemments to induce localgovernments to undertake sectoral expenditures thatare of national-ratherthan local-interest, andto use local governments as agents of national income redistribution policies. But reform is requiredboth to increase the effectiveness of transfers in achieving these sectoral and distributional objectivesand to reduce the adverse side effects of badly targeted or badly administered transfers.

xi. Perhaps the most important measure developing countries can take is to reduce theunnecessary adverse side effects of existing transfer programs-to reduce the uncertainty and

3

bargaining that now accompanies intergovernmental financial flows and remove incentives forstrategic behavior. But countries also need to expand transfer programs to adequately finance theexpenditures they have assigned to local government-particularly if these include primary educa-tion and preventive health car-and to target them more effectively.

xii. Reform is also needed in the arrangements by which municipal govemments obtainaccess to financing for capital investment. To an extent, improvements in allocation can be achievedby improving the targeting of grant programs-particularly where the preconditions for allocationby lending do not exist. But there is also a case for replacing grant financing with loan financing.Municipal credit institutions (MCls) are an increasingly popular means of administering suchprograms. The perfonnance of these organizations is mixed, however, and while some legal andorganizational steps may enhance their viability, experience suggests that what matters most is thegovernment's commitment to the MCI's independence and its willingness to provide a supportivefinancial environment.

Balancing regulation and electoral accowntabiliy

xiii. How much a system of urban service delivery should rely on accountability upward tocentral government-through regulation-or downward to constituents-through political partici-pation-is not an issue that lends itself to universal prescription. Neither extreme is advisable.

xiv. Some degree of accountability to central government through a national regulatoryframework appears to be appropriate to any structure of urban service delivery. Central regulationis clearly appropriate where local government behavior can affect national monetary, trade, or fiscalpolicy. It is also clearly appropriate where local governments are carrying out functions on behalf ofcentral government. But where impact of local government behavior is largely localized, andregulation requires detailed knowledge of local conditions and priorities, the case for centralregulation is more difficult to justify.

xv. The counterpart to central regulation is local political accountability-Ahe reliance onvoters to regulate the behavior of their political leaders. The view that local elections performimpeccably in this role does not stand up well to scrutiny. The validity of elections after long periodsof authoritarian rule appears particularly questionable.The advent of local democracy, whileincreasingly common, is therefore no panacea. There is nevertheless some evidence that specificchanges in election rules can influence the degree to which local elections function as referenda onlocal government performance. And there are altemative, and supplementary, means of holding localmunicipal leaders accountable.

Synchronization

xvi. While there is clearly no one way to organize the delivery of urban services, what doesappear evident is that the various pieces of the intergovernmental relationship have to fit together.This has become increasingly evident in countries that are undergoing political decentralization. Thepolitical impetus behind decentralization has prompted central governments to make politicalconcessions hastily. But granting local elections is a step that can be taken quickly. What is slow and

4

difficult is thieworkingthroughof newregulatory relationshipsbetweencentralgovemmentand localgovernment; the conversion of what had been annual budgetary transfers within a central govemmentinto intergovernmenlal transfers that arc transparcnt and predictablc, and the developmcnt of crediblelocal political systems. Many of the problems associated with the current wave of decentralizationarisc from the failure tomatch Lhe pace of political decentralization to the pace ofregulatory and otherorganizational reforms.

L INTRODUCTION

The Context

1.1 I Over the past 25 years, the developing world has been transformed from a world of ruralvillages to a world of cities and towns. In 1965, less than one quarter of the population of low- andmiddle-income countries was urban. By 1990, that proportion had increased to roughly half. Theurbanization of the developing world is a global phenomenon. As shown in Table 1, all the majorregions of the world have experienced dramatic increases in their degree of urbanization in the last =quarter century. By the year 2025, it is estimated that more than two thirds of the population of Xdeveloping countries will be urban.

Table 1. Trends in world urbanizaflom

Region % ubun 1965 % urban 1990

Sub-Saharan Africa 14% 29%East Asia 19 50South Asia 18 26Europe 40 60MENA 35 51Latin America 53 71

Source: 1992 World Development Report

i.2 National economies are also increasingly urban. Although figures on the geographicorigin of GDP are difficult to obtain, sectoral data is indicative of the importance of cities in nationaleconomies. In 1990, more than 80 percent of the GDP of the low- and middle-income countries wasproduced in the nonagricultural sectors-largely in manufcuring and services. And data on thesectoral composition of economic growth suggests that the urban sector is a principal engine ofdevelopment As shown in Table 2, growth ratsin manufacuring and services consistently exceededthose in agriculture in virtually all regions 6tfie world during the last decade. Even in regionsundergoing severe adjustment during the 1980s, growth rates in manufacturing and services farexceeded those in agricultre.

Table 2. Sectoral compoion of economic growth(Averge annual GDP growth, 198-90)

Region Agriculture Manufacturing Services

Sub-Saharan Africa 2.1% 3.1%. 2.5%

East Asia 4.$ 12.4 8.0South Asia 3.0 6.8 6.3

Europe 1.0 2.7MENA 4.3 3.4 1.9Latin America 1.0 1.7 1.7

Source: 1992 World Development Report

6

1.3 The demographic growth of cities and their contribution to national economies suggestthat cities work. Urban population growth is fueled by the prospects forjobs and higher incomes. Theavailable evidence suggests that these expectations have largely been met. The personal income ofrural-to-urban migrants is consistently higher in the place of destination than in the place of origin.Higher incomes in cities in turn reflect the greater productivity of labor in cities: studies of laborproductivity suggest that urban-rural wage differentials largely reflect spatial variations in laborproductivity.

1.4 Yet citido not deliver on the promise of a better quality of life to the extent that theycould. Despite the relatively high incomes of urban populations, thequality of services inmajorcitiesis poor. At least 170 million people in urban areas lack a source of potable water near their homes,and in many cases, the water that is supplied to those who have access is polluted ( World Bank 1992).Nearly 350 million people in urban areas lack access to basic sanitation (a figure that is believed tohave increased in absolute terms over the last decade (World Bank 1992)). In many large cities indeveloping countries, less than 70 percent of municipal solid wastes are collected and only 50 percentof households served (Bartone, Bernstein, and Wright 1990). In Mexico City, it is estimated that theaverage time required for a journey to work is between 2.5 and 3.5 hours (Legorreta 1989). Whiledata on the coverage of education and health services in urban areas are not available, the aggregatestatistics for Third World countries are disturbing: in half of low-income countries, fewer than halfof school age children are enrolled in primary schools (World Bank 1990a).

1.5 These service failures have wider economic and distributional implications. ChristineKessides' discussion paper notes that a lack of access to, or unreliability of, infrastructure servicescan have adverse effects on growth, forcing firms to seek higher cost altematives, which may in turnhave unfavorable impacts on profits and levels of production and consequently on investment andjob growth (Kessides 1993). A 1988 study of manufacturing establishments in Nigeria found thatcosts of private substitutes for public infrastructure services (generators, boreholes, vehicles forpersonnel and freight transport, and radio communications equipment) constituted 15 percent of thetotal machinery and equipment costs of small firms, and 15 percent of those costs in large firns.

1.6 Service failures also have distributional implications. The economic benefits of urban-ization have not been uniformly distributed- As countries have urbanized, poverty has urbanized aswell. It is estimated that by the end of this century, 90 percent of the absolute poor in Latin Americawill be living in cities, as will about 40 percent of the poorest in Africa and 45 percent of thosein Asia --(Rondinelli 1990). Where the formal system of services delivery fails to reach the poor, thesehouseholds must resort to alternatives that ofLen imply not only lower quality but higher costs. In theabsence of piped water supply systems, for example, households are forced to rely on purchases fromwater vendors. A survey of water vending in 16 LDC cities demonstrated that the unit cost of watersold by vendors is, on average, 12 times higher than that of piped water systems (World Bank 1992).

7

The Management Problem

1.7 Failures in the coverage and quality of services reflect, in part, aggregate resourceconstraints. The ability of an economy to provide convenient, reliable urban services is constrainedby the demands of other fundamental needs-food, clothing, basic shelter, security-in extremelypoor countries. But the available evidence suggests that the constraint on improved service deliveryis not merely one of resources. The level of resources already devoted to urban service delivery issubstantial. There is evidence, for example, that in the absence of conventional service deliverysystems, households commonly resort to more expensive altemative sources. In Jakarta, Indonesia,800,000 households have installed septic tanks, at a cost equal to three times the amount that wouldhave been required to provide connections to piped sewerage systems. Moreover, there is evidencethat governments-particularly central govemments-already spend a significant amount on urbanservices. While statistics on the sectoral and geographical pattern of government expenditure arescarce (panticularty--ih developing countries), estimates of the annual level of central governmentexpenditure on urban services' in the developing and transitional economies range from $100 to $200billion-approximately 2.5 percent to 5 percent of GDP.

1.8 The deficiencies in urban services in the cities of developing countries are therefore notmerely a reflection of absolute resource constraints, but also of problems of management Definingmanagement in a policy-relevant manner is difficult. Management analysis is an evolving field inwhich even the definition of the relevant variables continues to expand. While the traditional focusof management has been the application of good technique-accounting, organization planning,financial analysis-recent frustration with this approach2 has prompted a change in strategy; a shiftin the focus of analysis from internal procedural issues to the incentive structure; to the variousrewards and penalties confronting the individuals involved in the delivery of urban services. In partthis represents ashift in the wayinternal operations of an organization are addressed, achangein focusfrom administative procedures to the factors motivating individuals within the organization. But ithas also directed attention to the role played by individuals outside t-he organization: to the consumerinterest groups, unions, and central government regulators whose behwa.ior also influences theperformance of service delivery systems ( Ostrom, Schroeder, and Wynne 1993)?3 While this

Box 1. A typology of decentralization

D. Rondinelli's classic typology identifies four different categories of decentralization. All represent transfers ofpower from centrl government administration. The typology distinguishes types of decentralization on the basisof the organization to which power is transferred:

Deconcentration is defined as a transfer of power to local administrative offices of the central government;Deleganon as the transfer of power to parastatals;Devolution as the transfer of power to subnational political entities; andPrivauization as the transfer of power (and responsibility) to private entities.

1. Defined as water supply, sewerage, intn-city roads, drainage, subsides to urban mass transit, primary education andhealth.

2. The inability of the traditional analytic framework to yield effecdve normative conclusions is summarized in Alsrael'sInstitutional Development.

3. While not denying the importance of administrative procedures, this view would argue that procedural reforms areonly effective if the incentive structure is supportive; that what matters in the first instance, for example, is not accountingsystem, but the motivation of the staff to use iL

8

approach seems to promise a more robust model of organizational behavior, it makes for acomplicated analysis. It argues that to understand organizational results, one must understand themotivations of the myriad of individuals who play some role in the provision of urban services.

1.9 One product of this perspective is the view that problems in service delivery in part arisefrom the perverse relationships between political leaders at the central and local levels and theirconstituents. Critics of thepresent,often centralized, structureofgovernment indeveloping countriesargue that such structures are inherently incapable of responsive administration. Because theconcerns of central govemment become increasingly predominant as the locus of decisionrmakingmoves away from beneficiaries, because the costs of influence become increasing high as decisionsare centralized, and because the quality of information about local conditions becomes increasinglydistorted as it moves from field officers to central administration, centralized political systems are(according to this line of reasoning) inherently unresponsive. Rondinelli, for example, argues that"rarely do incentives exist for central government ministries to perceive citizens as theirclientele"(Rondinelli 1990).

1.10 Whatever the merits of these arguments, they were largely of academic interest until themid 1980s. The internal political structures of most developing countries were stable. Centralized,often authoritarian regirres, prevailed. Whatever the merits of decentralization, existing intergovem-mental relationships appeared to be immutable. But during the 1980s, the situation began to change.Political decentralization is now, in fact, a very widespread phenomenon. Out of the 75 developingand transitional countries with populations greaterthan 5 million (WorldBank 1992). all but 12claimto be embarked on some form of transfer of political power to local units of goverment. The formand extent of decentralization varies. In parts of Africa, national govemments are creating localpolitical entities in territories chat were formerly solely under the administration of central govern-ment. In Eastern Europe, what were formerly local administrative units of the central governmenthave been transformed into separate political entities, with leadership chosen by local election ratherthan by appointment through the party structure. In Latin America, decentralization has meant a shiftfrom centrally appointed mayors to mayors chosen by election. (in some Latin American and Africancountres, the most significant aspect of decentralization is central governments' tolerance of localvictories by opposition parties.) In some countries, particularly in Latin America, political decentrali-zation has been accompanied by increases in revenue sharing or by nominal changes in the legaldefinition of local functional responsibilities.

1.11 The motivations for decentralization-as it is observed in practice-appear onlytangentially related to administrative reform. Political analysts have, in fact, suggested thatdecentralization stems from a more fundamental cause: the need of national political leaders toaccommodate or deflect increasingly strident demands for power sharing by groups that havetraditionally been excluded from it. Prud'homme, for example (referring to the Third World ingeneral), has characterized decentralization as "a political strategy by ruling elites to retain most oftheir power by relinquishing some of it" (Prud'homme 1992). Davey, referring to Africa, hascharacterized decentralization as an attempt by bankrupt central governments to create anew targetfor political dissatisfacion (again without relinquishing real power). In Eastern Europe, where theconcessions came too late, decentralization was (again in Prud'homn,e's words) "a hasty effort bynewly victorious political forces to consolidate their positions" (Prud'homme 1993). Variousanalysts have speculated on the underlying causes of this rise in the stridency of political opposition.

9

It has been attributed to the conspicuous economic failure of the centralized state (in the ex-Easternbloc, but also in Africa) (Hall 1993), to the relative absence of war and civil unrest (with theconsequent decline in acceptance of strong authoritarian government) (Potter 1993), and to theemergence ofeducatedurban middle classes-and the consequentdecline of traditional patron-clientrelationships between the govemment and governed (Fukuyama 1992). Whatever the underlyingcause, it is clear that the decentralization now occurring is not a carefully designed sequence ofreforms aimed at improving public sector performance. It more often takes the form of a reluctantand disorderly seriesofconcessions bycentral governmentsattemptingtomaintain political stability.

1.12 As a result, while decentralization has brought change in the structure of urban servicedelivery, it has not necessarily brought improvement. Unfortunate administrative consequences ofdecentralization are evident in a variety of contexts:

* In Hungary, the concession of local political autonomy preceded the separationof local budgets from the central govemmentbudgeting system. As aresult, localgovemments lack a financially sustainable revenue base and have been forcedto sell assets to cover recurrent costs (raising the prospect that the social safetynet-for which local governments bear major responsibility-will run out offunds in the near future).

* In Brazil, where decentralization was reportedly motivated by the politicalrepudiation of two decades of military rule, decentralization has taken the formof a substantial increase in revenue sharing and in the taxing powers of localgovernment. It was not, however, accompanied by a corresponding delineationof local expenditure responsibilities. Thus, although local governments havemore money to spend, they are no more accountable for the quality of theirservices than they were before the reforms.

- In Ghana, decentralization appears to be merely the attempt of a beleagueredcentral-govremment topush the political (and financial) costs ofgovernmentontoaltemative levels of govemment, widtout actually relinquishing power (inter-view with Ken Davey). Political decentralization has not, so far, been accompa-nied by a commensurate decentralization of authority: the creation of locallyelected legislative bodies has not been accompanied by a transfer of significantdecisionmaking authority to local government. The Ghanaian governmentcontinues to appoint the municipal executive and the heads of municipaldepartments and effectively control local spending decisions.

1.13 Despite its political dimensions, decentralization still presents reformers-both domes-tic and in the donor community-with an opportunity to promote administrative reform. Politicalpressure for decentralization has resulted in a high degree of fluidity in the structure of the publicsector. With govemrments forced to make fundamental changes in the structure of the public sector,it presents an opportunity to promote the kinds of fundamental changes in intergovemmentalrelations that have tended to be frustrated in the past. This opportunity is presumably a transitory one,and one that the should be grasped.-

1.14 The potential impacts of such reforms may extend beyond sectors that are consideredurban. Changes in the allocation of power between levels of govermnent have implications fornational efforts to reduce poverty and manage the macro economy (particularly to the extent they

10

reduce central control over the instruments of fiscal policy (Prud' homme 1993)). Decentralizationalso affects the delivery of social services-primary education and health-as well as nonurbaninfrastructure. Decentralization is in fact already the subject of World Bank sector work in education,transport, and health.

1.15 There is no road map here. The world now consists of more than 200 ongoingexperiments. Of the countries that have embarked on decentralization, none yet has the length ofhistory that would permit an expost evaluation. The developed, Western economies-which mightbe assumed to offer models-have themselves adopted different approaches to the organization oftheir subnational governments and several are in a constant state of experimentation. The objectiveof this paper is therefore modest to document past approaches to the development of (urban) localgovernments, to derive lessons from that experience, to assess the implications of decentralizationfor these efforts, and to propose an approach (derived from both theory and such experience as exists)that might be used in the future.

IL A FRAMEWORK FOR ANALYSIS

Defining the Problem

The problem iS not municipal government

2.1 Much of the early work on urban service delivery systems-particularly by the donorcommunity-focused on the internal administration of municipal government. Urban servicefailures were defined as a problem of internal administration-specifically a problem of poor taxadministration, poor accounting, and poor capital investment budgeting. As described in a recentsurvey of institutional development components in World Bank urban projects (Haines 1992),approximately two thirds of the identifiable targets of World Bank municipal development projectsduring this period were aimed at local taxation, accounting, or capital budgeting. This pattern wascharacteristic of other multilateral aid organizations and bilateral donors.

2.2 In retrospect, this now appears to be too narrow a definition of the problem. First, inmany developing countries, municipal governments did notexist in the sense that was often assumed.While units of local administration existed, these were not distinct political bodies, drawing theirlegitimacy from local-rather than national-political constituencies, and possessing significantautonomy. As a result, the assumption that local political leaders had either the motivation or theauthority to respond to their constituents in the same manner as their counterparts in industrialdemocracies was often misplaced.

2.3 Moreover, the assumption that municipal governments in developing countries hadclearly defined responsibilities-and that these were the functions typically assigned to localgovernments in the industrial world-was also often erroneous. Despite the provisions of nationalconstitutions, the sectoral responsibilities of municipal govemment were very limited and reflectedthe ad hoc accretion of past practices and political accommodations. Central governments (in federalcountries, large states) would control major capital investnent decisions in water supply, sewerage,road construction, or public education and health, leaving the municipal government responsible foroperations and maintenance of these services. In some countries, local govemment functions werelargely limited to the maintenance of public order and the cleaning of public streets. In Juarez,Mexico, for example, "strengthening local government" would be largely tantamrount to expandingthe municipal police force.'

2.4 The assumption that financthe lack of resources-was a key constraint was alsomisplaced. Although municipal governments were poor-both in terms of absolute revenues percapita and in terms of the share of total public expenditures (relative to their counterparts in industrialcountries), local government functional responsibilities were correspondingly limited. Moreover,given the absence of clearly defined local responsibilities, little assurance existed that increases inlocal revenues would be devoted to functions that (in the view of donors) represented high priorities.

1. In 1988, nearly half cf Juarez' budget (excluding transfers to the municipally-owned water parastatal) was, directlyor indirectly, devoted to the policc.

12

2.5 Finally, the assumption that problems in internal administration could be addressedthrough the transfer of technical knowledge was also misplaced. Donors tended to operate as if theconstraints on administrative improvement were constraints of knowledge. Thus technical advice onproperty valuation techniques were prescribed as the solution to low property tax yields; technicaladvisory services were similarly prescribed as the solution to problems in accounting and capitalbudgeting. Subsequent evaluation reports and sector work, however, suggest that technical advisorswere often marginalized and their advice ignored; and that problems in internal administration werenot due to a lack of knowledge but instead reflected rational responses to firmly established, ifperverse, institutional incentives.

The problem is the system of urban service delivery

2.6 The conclusion that emerges from this experience is clear. Urban service deliveryappears to be a problem that cannot be addressed by taking the organizational context as a given andattempting to change the behavior of one organization-municipal government-within it. Instead,it appears to be a problem of the public sector as a whole, and one that must be addressed by lookingat the variety of factors that influence the performance of the public sector and those factors'implications for urban service delivery. Constraints on urban service delivery are not only the levelof the property tax or the sophistication of the accounting system, but also the broader set of perverseincentives embedded in the relationship between central and local governments. In failing to addressthe incentive structure, reform efforts not only undermine the prospects for achieving administrativetargets; they also appear, in retrospect, to be the most significant constraints on the performance ofurban service delivery systems (World Bank 1993a)3'

2.7 Efficiency and responsiveness. In a sense, the problem of urban service delivery is bestdefined in terms of characteristics of the delivery system, rather than the problems of a discreteorganization. The objective ofreform is not "strong local government" but an efficient and responsivesystem of urban service delivery. Efficient in a technical sense-in the application of productiontechnology so as to produce the maximum amount of output from a fixed quantity of inputs-andefficient (or responsive) in an economic sense-in its terms of its effectiveness in producing the mixof products that corresponds to the varying preferences of consumers.3 The performance of a servicedelivery system can thus be evaluated both in terms of its efficiency in the transformation of inputsinto products and of its efficiency in producing the mix of products that matches the effective demandof consumers.

2.8 Equity and economicstability. Efficiency isnotthesolecriterion by whichamunicipalservice delivery system should be evaluated. While technical efficiency and responsiveness may bethe paramount objectives, municipal service delivery cannot be isolated from other objectives of the

2I These conclusions are based on a 100 percent survey of all OED project audits and PCRs on urban developmentprojects through December 31, 1992. (Sce table in Annex). It should nevertheless be noted that (I) OED audits covera early period in Bank interventions; and therefore do not reflect the impact of changes in the Bank's approach, if any;and (2) tlat the tiine frame covered by the audit (the disbursement period) is often too short to permit an assessment ofthe susmainability of a intervention's impact

3. In the decentralization literature, the two categories are variously referred to as "production efficiency and demandefricieicy" (Campbell, Peterson); or efficiency and efTectiveness (Rondinelli, Ostrom).

13

public sector. Decisions on the level and financing of municipal services have important distribu-tional implications: expenditures on education and health, forexample, are often thc mostsignificantforms of income transfers in developing countries. Moreover, to the extent that central governmentseffectively concede control over fiscal policy to local governments, certain financing arrangementscan have adverse implications for the stability of the macroeconomy. The implications of a givenstructure of service delivery for poverty reduction and macroeconomic stability are therefore alsoimportant considerations.

Defining the Instruments

Public economics perspecdve

2.9 What are the means by which these objectives are to be achieved? Several academictraditions contribute different parts of the answer to this question. The public economics literatureprovides a useful organizing framework. Public economics sees this as an assignment issue: aquestion of dividing functional responsibilities between levels of government, and allocatingcorresponding revenue sources. In the classic formulation (Musgrave and Musgrave 1984) the publiceconomics literature assigns three roles to the public sector as a whole: (a) macro stabilization-theprovision of a stable economic environment for the private sector, (b) income redistribution-thereduction of poverty, and (c) resource allocation-influencing the allocation of resources wheremarkets fail to do so efficiently. The public economics model conventionally assigns the first two ofthese roles to central government. Central government has responsibility for stabilization policy,largely on the grounds that local economies are too open to permit countercyclical measures to beimplemented effectively. Income redistribution policy, similarly, is considered to be the prerogativeof central government, on the grounds that local attempts to address income disparities are likely toinduce inefficient migration. Subnational governments enter the picture only with respect to the thirdrole of government-resource allocation. Noting that tastes and preferences for public services varyamong populations, this framework argues that if the benefits of particular services are largelyconfined to local jurisdictions, welfare gains can be achieved by permitting the level and mix of suchservices to vary accordingly. If local consumers are confronted with the costs of alternative levels ofservice (according to this viewpoint), constituents can be forced to reveal their preferences (throughvoting for rival political candidates, voting on tax initiatives, or similar means). In this respect, localpolitics can approximate the efficiencies of a market in the allocation of these local public services;L4pricing" municipal services and relying upon the local political process to clear the market.

2.10 The organizational implications of the public economics framework are straighnfor-ward. Responsibility for discrete public services should be assigned to the level of government whoseboundaries incorporate the affected beneficiaries. That level of govenment should then be assigned aconrespondingpricing instrument-ataxwithacorresponding inciderce-with which to ascertain demand.

Institudonal analysis perspective

2.11 Although useful as a starting point, the public economics framework is not an agendafor action. In effect, it assumes away certain behavioral and technical factors that appear to be

14

important. It assumes that local politicians and the people who work for them are solely interestedin serving the interests of their constituents, and that constituents have both the means and theincentives to make their interests known. The public economics framework also assumes awaycertain practical considerations. It ignores the high administrative costs that would be associated withassigning each function of the public sector to a distinct unit of government with its own fieldadministration. And it ignores economiesofscale in serviceproduction and thus the potential conflictbetween an organizational solution that would be most responsive, and one that would be mostefficient in a technical sense.

2.12 The behavioral implications of alternative organizational arrangements are addressed inwhat has been called the institutional analysis perspective. This is a framework that explains thebehavior of politicians not as altruism, but as the rational pursuit of self-interest-the behaviorsrequired for advancement in a political career-and where the behavior of bureaucrats is similarlyexplained as a rational, self-interested response to the working rules that allocate rewards and costswithin the organization. As Ostromargues,"individuals-politicians andbureaucrats-re boundedlyrational and make choices according to their preferences. Individuals are not assumed a priori to beencapsulated by their organizations, nor to find organizationally rational behavior necessarilyconsistent with individually rational behavior" (Ostrom, Schroeder, Wynne 1992).

2.13 Proceeding from this framework, the institutional economics literature argues that theway to change the performance of an organization is to change the incentives of the people who workin it. One of the normative conclusions arising from this field is its advocacy of organizational rulesthat would render service providers more vulnerable to client groups. This is elegantly summarizedin Hirshman's phrase "exit and voice," in which he argues for organizational arrangements that willoffer consumers the option of either taking their business elsewhere (exit), or increasing theirinfluence overproviders through "voice"-enhancing the lines of communication between providersand constituents-through fonnal electoral processes and alternative means of articulating prefer-ences, including the use of local taxes and charges as a means of testing beneficiaries' willingnessto pay.4

2.14 Other contributors to this field have noted the preconditions that affect the articulationof voice. Israel has contributed the term "specificity" to this glossary, noting that constituents havegreater success articulating voice when the entity to which they are addressing thnmselves has a clearlydefined output for which it is responsible (lsrael 1987). Ostrom has emphasized the importance ofminimizing transaction costs in the articulation of voice, noting dtat constituents are unlikely to participateunless they perceive the expected benefits to exceed the costs (Ostrom et. al 1992).

2.15 The normative prescriptions arising from this framework are compatible with theframework p.oposed by public economics. (Ostrom in fact incorporates the public econompicframework into her nornative model of "polycentric institutional arrangements".) What the

4. The literature on "exit" has been extensively covered elsewher- and will not be further examined here

15

institutional analysis literature contributes is a more extensive list of policy handles: a view that theperformance of organizations is not only determined by their nominal powers and responsibilities,but by the motivations of the people who work within them. This directs attention to such issues aswhether the mayor is appointed or elected (and what sort of career trajectory confronts a successful-or unsuccessful mayor), and the mechanisms by which interest groups can make their wishes knownto local governments and exert pressure to have them fulfilled.

Administrative distinctions

2.16 A third strain in the literature addresses-or at least provides a framework foraddressing-the technical problems raised but not addressed in the public economics framework.The literature has long recognized a distinction between the "provision" and "production" aspectsof service delivery-a distinction between the role of deciding how much is to be produced and therole of producing it (Musgrave and Musgrave 1984). This distinction is useful in addressing theapparent conflict between the demand-responsiveness of decentralized decisionmaling and thepresumed economies of scale of centralized production. As Silverman, among others, has recentlynoted, production can be organizationally separated fromprovision; theentity thatdecides how muchto produce need notbe the entity that undertakes the production (Silverman 1992). Such a separationallows the responsiveness benefits of local decisionmaking to be captured by assigning the provisionrole to local government, while allowing economies of scale to be captured by permitting largerentities of govermment (or private firms) to be responsible for production.

2.17 A parallel distinction is useful in addressing the high administrative costs that wouldresultfrom assigning each function of governmentto aseparatepublic entity.AsCampbell has noted,a single tier of government can act in muldple capacities. The local government can act as both theprovider of local public services-clearing the market in local public goods in response to localexpressions of demand-and as an administrative agent of higher levels of government (Campbell1992). In this respect, a single level of government can assume functional responsibility for multipleservices, thus reducing the aggregate administrative costs of the public sector.

2.18 Evidence from the industrial countries suggests that these distinctions are not merelyacademic. The industrial countries (which may be assumed to be relatively successful in this regard)all have complex structures of service delivery in which these distinctions play arole. The separationof provision from production is commonly observed in the G5 countries, particularly in the execu-uonof capital wo:*s. Although local governments play a large role in the delivery of infrastructureservices in the G5 countries, little construction is done by municipal employees. Major offPsite capitalworks are typically constructed by private contracting firms. Residential infrastructure in developedcountries is almost entirely produced by private-sector developers (prodded by local developmentregulations.) The production of such services as water supply (in the case of France and the UnitedKingdom), solid waste management (in the U.S.) and bus transit is also assigned to private firms inthe G5. Similarly, the use of local governments in an agency role is widespread in the G5 countries.In the U.S., for example, the federal government relies on local government to implement some ofits major income-distribution policies, including the administration of AFDC (Aid to Families withDependent Children) and food stamps.

16

2.19 Overall, the conclusion of this literature is that there is no single institutional arrange-ment that can be universally prescribed for the delivery of urban services, that what is important arenot the organizational labels, but rather the relationships-the rules that govem the transactionsbetween national government political leaders, local govemment leaders, staff, and consumers. A"good" arrangement is likely to be a very complicated one and one that is not defined merely by thedesignation of municipal responsibilities and revenue sources. It is one in which the interests ofnational political bodies and local interest groups are incorporated in different ways, with municipalgovernments performing as agents of central govemment in some capacities, and as independentdecisionmaking bodies in others. But while the system can be organized in a variety of ways so asto yield a relatively efficient, responsive structure of governance, not all such arrangements workequally well. Evidence from developing countries illustrates a variety of suboptimal arrangements,as discussed below.

A Typology of Intergovernmental Relations

2.20 The intergovernmental relations of a country are difficult to know. The official rules-the legal framework-are at best only an indicator of the actual locus of decisionmaking power andthe influences that bear on decisionmakers. It is useful nevertheless to attempt to articulate commonpatterns in intergovernmental relations, to illustrate the variety of circumstances that exist.

2.21 Three patterns are discemable. The first might be called the overcontrolled local sector,where subnational governments are in effect merely administrative arms of the central government.Its obverse is the undercontrolled local sector, in which each tierof subnational governmentis almostsovereign and competes with other levels of government. The tiird minght be called the perverselyregulated local sector, where local governments have some degree of political autonomy but wherethe relationship with central political leadership is characterized by perverse incentives.

The overcontroUed local sector

222 This is historically the classic pattern in developing countries. It has two typicalcharacteristics. First, avery high proportion of total public expenditure is made directly by the centralgovernment (or in federal countries, by large states). Second, local government, even within itscircumscribed expenditure role, functions largely as an administrative arn of the central government,with the central government appointing the municipal executive, and dictating virtually all expen-diture and revenue decisions.

2.23 Central government's dominance of total public sector expenditure is the commonpattern in developing countries. As shown in Figure 1, the local share of total public sectorexpenditure in developing countries-even large countries like Mexico-is well below 10 percent.This is in contrast to the G5 countries, where the local share ranges from 17 percent (France) to 33percent (Japan).5

5. These statistics, based on the IMF Government Finance Statistics Yearbook, are at best crude indicators of the degreeof centralization in these countries. As noted later in the text, they do not, for example, rellect the degree to which localexpenditures arc controlled by central governmenL Certain statistical conventions and data constraints in the IMF data

17

Figure 1. Local share of public expenditure

JapanU.K.U.S. -- -

Germany _ _ _ _ _France * i -

Czech i nHungary 1Romania =nhm

KoreaUColombia

Chile UIEISuPhilippines

ThailandTurkeyBrazil

JordanTunisiaMexico

a 5 10 15 20 25 30 35 40

2.24 Local govemments in these countries, moreover, bear no resemblance to autonomouspolitical entities envisioned in the public finance literature. To begin with, central governments,rather than local voters, choose the municipal political executive. (Elected councils are oftenpernmitted, but function in a purelv advisory role.) As shown in Table 3, direct appointment ofmunicipal executives is still the pattern in many large LDC cities. Elsewhere, central control overlocal political outcomes is discretionary but is widely used. In India, for example, state govemmentspossess the legal authority to dismiss mayors whose performance they find unsatisfactory. Sinceindependence, at any given time, 40-50 percent of the local authorities have been under statesupersession. Bombay's current mayor is, for example, a state appointee.6

also tend to understate the degree of public service decentralization in the G5 countries, and the consequent contrastbetween 05 and developing country patterns. First, the IMF data do not include the self-financed expenditures of publicutilities (recording instead only tax-financed subsidies to parastatals). To the extent local public utilities in industrialcountries are largely self financing (and those in developing countries are not) this convention would tend to understatethe difference between the two groups of countries. Second, IMF data do not pennit the disaggregation of governmentexpenditure by function (except for the industrial countries and a small number of developing countries.) As a result.it is not possible to identify the proportion of central government expenditure in urban service delivery. It can be surmisedthat such a comparison would indicate that the local share of urban service expenditure in industrial countries is muchhigher than in developing countries. IMF statistics do indicate, for example, that in the industrial countries, localgovemments account for the 40-60 percent of public expenditure on capital investment, which anecdotal evidence wouldsuggest is not the case in developing countries. In the 05 countries, central government expenditures are dominated bydirect transfers to individuals-essentially social safety net expenditures-whereas in developing countries (judgingfrom a small sample), a larger proportion of central govemment expenditure is devoted to capital works. If data on theaggregate expenditure on urban services (including self-financed utilities) were available for both developing anddeveloped countries, it can be surmised that the contrast in this indicator of decentralization would be greater than isshown by the available statistics.

6. Recent legislation has restricted, but not eliminated, states' power to supersede local elected officials.

18

Table 3. Structure of poliilcal accountability In major LDC cities

City Mayor Council

Bonmbiay appolnied b)y state teonpomrily dissolved

Jakarta appointed by govemment directly eleccld,no legi9lative ptower

Mcxico D.F. appointed by government directly clecied,no legislative power

Sao Paulo directly elected directly ciccied at large,legislativc power

Seoul appointed by government directly elected,no lcgislativc power

Budapcst directly clected ciectced, 114 directly by district,3/4 aT large by party list

Lagos dircctly clectcd clected

Shanghai clected by council ciceted from one-party slate

2.25 Central governments alsodirectly control the allocation of municipal expenditure in thisgroup of countries. Personnel is typically the largest single item of local government expenditure, andthe ability of local govemment to recruit, retain, and motivate staff is critical to its ability to providemunicipal services efficiently. In many developing countries, control over local personnel manage-ment decisions rests with central government. Central governments often control the number ofpositions local governments are allowed to maintain at each grade, the starting salaries and paydifferentials between grades; the level of cost of living increases; and the appeals process fordismissal. In Turkey, forexample, the staff list of each municipality isfixed by the government, alongwith the corresponding salary scale. Any amendments to the staff list have to be approved by thecentral government in a laborious process involving the Ministry of Interior, the State PersonnelOrganization, and the Council of Ministers. In some countries, central governments are directlyinvolved in individual recruitment and promotion decisions. In Ghana, for example, local govern-ment staff are directly recruited, promoted, and paid by sectoral ministries of the central government.In the Philippines, municipal treasurers and tax assessors (until recent reforms) were directlyrecruited, supervised, and paid by the central government ministry of finance. In Indonesia, all full-time staffing positions are subject to central government recruitment and promotion and are paiddirectly by central govremment.

2.26 Central governments also control the sectoral composition and size of local govemmentbudgets. In Morocco, for example, each municipality's budget must be approved by the Ministiy ofInterior before the funds can be disbursed against it-an approval process that includes a verificationthat all centrally mandated expenditures are in the budget and a verification that personnelexpenditure aggregates agree with the Loi de Cadres for each year, the payroll of authorized gradesand positions and with the Treasury allocation for municipal personnel expenditures. In Senegal,

19

similarly, the annual budget review process includes a line-by-line negotiation of the expenditureestimatesofeach local authority. In the Philippines (also until recentreforms),theoverall sizeof eachlocal govemnment's budget was determined by the local representative of the finance ministry. Inad-ition, central budget regulations required that municipalities allocate 20 percent of their revenuesharing for development projects approved by the government, 18 percent of their general fund forthe national police, and 5 percent for aid to hospitals within the province; and thatmunicipalities limitthe proportion of the budget spent on personnel to 45 percent. (One study estimated that an averageof 46 percent of all local expenditures in the Philippines weremade undercentral govemmentmandate.)

2.27 Municipal revenue levels are also regulated. In five of the eightLDCcities listed in Table4, forexample, local governments have no discretion overtherate of theirprincipal tax. Governmentsuse this control to keep the rates of local taxes extremely low. In Jakarta, for example, the governmentof Indonesia limits the property tax rate to 0.1 percent of assessed value. Even where localgovernments have snme nominal discretion over the rate, central governments control the factors thatdetermine how much these sources yield. In the Philippines, municipal governments are permittedto impose a rate on the property tax between 0.5 percent and 2 percent of assessed value-but theregulations governing assessments are such that the maximum effective rate is less than 0.1 percentof market value. In India, the combination of state-imposed rent control and Supreme Courtinterpretations of property tax legislation have resulted in the virtual exemption of all but the city'smost recently constructed housing stock. In Mexico-in an environment in which annual inflationhas ranged as high as 150 percent-local governments are forbidden to adjust assessments forinflation unless they obtain the simultaneous concurrence of the mayor, the council, the governor,and the state legislature.

Table 4. Sources of tax revenue in major LDC cities

City Largest Local Tax Other Major Tax

% of tax % of taxName Rate control revenue Name revenue

Bombay octroi local 66% property 13%Budapest income central 95Jakarta automobile (1) Lentral 72 property IILagos property local 90+Mexico D.F. payroll central 58 property 21Sao Paulo services local 80 property 17Seoul property central 25 tobacco 21Shanghai business profit (2) central 50 industry 30

(1) includes separate taxes on purchase and ownership(2) amount of tax to be shared with central is negotiated

2.28 The performance of the overcontrolled system cannot be evaluated or fau!ted onempirical grounds. Some students of organization have in fact argued that the centralized, verticallyintegrated bureaucracy is in fact the most efficient form of organization2 But it can be argued thata system that concentrates such a large proportion of expenditure decisions in the hands of ministers

7. Max Wcber, a pioneer of modem management science, was an advocate of this point of view.

20

whose constituencies are national, and pernits so few channels of demand expression throughinfluence over local political leadership, would have difficulty responding to its constituents-particularly where the central government revenue structure is also centralized (with heavy relianceon indirect taxes and little use of user charges) and thus cut off from price-related indicators ofdemand.

2.29 Anecdotal evidence from political observers also tends to cast doubts on the centralizedsystem. As one observer of the local political scene has noted, "the practice of appointing municipalexecutives has often resulted in the mayor's position being held by individuals with only short-terminterests in municipal affairs. Where careers are dependent on the fortunes of political sponsors,officials aremorelikelytoadheretonationalpoliciesattheexpenseoflocaldccumsaes"(Lowder 1986).

2.30 There is also evidence that the administrative demands of this tight system of regulationoverloads the administrative abilities of the public sector in some developing countries. Centralgovernments thus find they lack the infonnation required to exercise this control intelligently. InTurkey, for example, governors are permitted one week to approve or modify a budget once it hasbeenpassed by themunicipal council. Approval isthereforevirtually automatic. The factthatrevenueestimates have been inflated unrealistically to satisfy the central budgetary regulations goesunexamined. In Kenya, similarly, the Ministry ofLocal Governmentisrequired to approve ormodifyall local budi'ets before the commencement of the fiscal year, but lacks the basic information (suchas data on actual income and expenditures of any previous year) on which to base its approvals. Underthese conditions, the overcentraized system is perhaps the worst of both worlds, in the sense that thereisneither aclearbur-aucraticchain of commandnoraclearlydefined scopt,oflocal discretion.Underthese conditions, central regulation merely obfuscates responsibility.

The undercontrolled local sector

2.31 The obverse case is the undercontroiled local sector, in whichi there are multiple levelsof govemment, each with political autonomy and autonomy over expenditure and revenue, butwithout any clearly defined functional responsibility. Brazil is per)haps the extreme example of thispattern. Brazilian mwzicipios have historically enjoyed complete political autonomy-with councilsand mayors chosen through competitive local elections (exceitduring a short interruption during themilitary regime of the 1960s). They have complete expenditure autonomy and freedom over the ratesof the taxes that have been assigned to them. When the extremely productive value-added tax wasintroduced (at the state level) in Brazil, the mwicipios were assigned a fixed 20 percent share of theproceeds, without restrictions on its use. No attempt is made, however, to define the functionalresponsibilities of each tier of government in Brazil. Brazilian legislation, while designating anextensive list of functions in which municipal government may choose to act, assigns the samefunctions concurrently to state government. (Only one function-natural gas-is exclusivelyassigned to the state level, and only one-urban transport-is assigned to municipalities.) As a result,both state and municipal governments may operate primary schools, health services, road construc-tion and maintenance programs, or any other public service simultaneously within the samejurisdiction. In Brazil, the defacto division of labor between levels of government reflects a patternof historical bargains and ongoing negotiations.

21

2.32 Again, the impact of this arrangement on the efficiency and responsiveness of urbanservice delivery has never been tested empirically. It would seem, nevertheless, to set up two kindsof perverse incentives. First, itwould seem to obscuretheaccountability of local governmentsto theirconstituents. Without clearly defined functional responsibilities, local government could not easilybe held responsible forthe outcomes of any particular service by their constituents. Second, itwouldseem to set up a perverse set of relationships between local government politicians and theircounterparts at the state (and central government) level. Without a clear distinction between thefunctions of each tier-in effect without a hard budget constraint on the extent of state participationin functions that a mayor is under pressure to provide-the extent of state participation is likely tobe determined through bargaining. Where state governmentresourcesare allocatedon this basis, theyare unlikely to reflect either the priorities of local consumers, nor those of the state government, butrather the lobbying skills of local politicians.

The perversely regulated local sector

2.33 While Brazil is an extreme example, it provides the polar case for the set of internediatecases that constitute the more prevalent pattem; countries in which there is some degree (defacto ifnot dejure) of local political autonomy, and where there is some (de jure) assignment of functionalresponsibilities, but.there is a built-in vertical gap-a lack of correspondence between the revenueauthority of local government and its expenditure responsibilitieshat is addressed in undesirableways.

2.34 There are two undesirable ways in which the vertical gap is closed. Perhaps the mostcommon is through direct, but ad hoc, expenditures by central government ministries. The practiceof direct intervention by central ministries is often the legacy of past crises. As Lowder describes it,central intervention is often provoked by disasters-outbreaks of communicable disease, forexample-and the perceived failure of local government to respond. "In response to these crises,specialized technical agencies are introduced to manage utilities such as potable water, electricity,housing for the poor...these are usually autonomous or responsible directly to a minister, andempowered to ovenride both the spheres of action and the territorial jurisdiction of other authorities"(Lowder 1986). In South Asia, the government's intervention tends to be permanent. Central (or inIndia, state) public works ministries and developmentauthorities takeonresponsibilityformunicipalcapital works, leaving only the task of operations and maintenance to local government. In LatinAmerica, the participation is more ad hoc. In Venezuela, for example, although municipal govern-ments have the legal responsibility to provide water supply, sewerage, urban roads, and powerdistribution, agencies of the central government intervene in the provision of these services at will(and in fact account for the vast majority of public sector expenditure on these services).

2.35 The pattern of open-ended participation by central government ministries would appearto set up the same perverse incentives as prevail in the under-controlled (Brazil ian) model: a structurein which constituents would have difficulty holding local governments accountable for any specificfunction, and where the mayorwould be encouragedto act as a lobbyistbefore the central govermmentministries, rather than as an individual ultimately responsible for specific functions. The more stableSouth Asian approach contributes an additional nuance. With responsibility for capital investment

22

assigned to state agencies, but operations and maintenance left to the municipal level, disputes overdebt liabilities and refusals to assume maintenance obligations on new assets are common. Given therevenue advantage that states have over local governments, overinvestment in capital (relative tooperation and maintenance) is also characteristic of South Asia.

2.36 Countries also fill the vertical gap through intergovemmental transfers. As described inChapter II1, transfers have an important role in virtually any system of multitiered govemmentfinance. The problem is not transfers, but the terms and conditions on which transfers are provided.Intergovernmental transfers take various undesirable forms. First, in some countries, the level oftransfers is itself unpredictable or largely determined by negotiation. In Kenya, for example, thecentral government is legally obligated to make an annual payment to local government in lieu ofproperty taxes owed on govemment-owned property. It has often failed to do so, a failure that hasset off a cycle of mutual default, such that local governments no longer service their debt to thegovernment loans authority, the government hospitals fail to pay their bills to the municipal waterauthorities, and the water authorities refuse to pay their income taxes to the government.

2.37 Even where recurrent transfers are distributed according to formula, transfers may haveperverse incentives effects. Dearness allowances in some Indian states and the SDO grants inIndonesia both fund part or all of local personnel costs, for example. This encourages localgovernments to lobby for more positions, regardless of need. Transfers based on the size of revenuegaps similarly encourage municipal governments to exaggerate expenditures or reduce local taxeffort. In Morocco, for example, the size of a municipality's transfer is based on the gap between itsestimated revenues and expenditures. Although both revenue and expenditure estimates must beapproved by govermment-limiting the scope for local strategic behavior-debt service is automati-cally included as an element in approved expenditure. Even unconditional transfers, such as Brazil's,may embody arbitrary interjurisdictional subsidies. Brazil's largest transfer, for example, is distrib-uted on the basis of the "origin" of value-added tax collections. But the value-added tax is collectedatthe point of production-from major manufacturing enterprises-not at the point of consumptionwhere much of the incidence presumably falls. The result is a cross subsidy from consumersthroughout Brazil to the residents of industrial enclaves-a transfer that, for example, enables themwnicipio of Sao Bemardo (the site of Volkswagen do Brasil) to operate a municipal symphony,courtesy of Volkswagen buyers throughout the country.

2.38 In developing countries, transfers are also used to fmance capital investment. And again,while this may be an appropriate roifor central government (particularly where long-term capitalmarkets are not well developed), the allocation mechanism for capital financing is one that tends tobe more responsive to the political interests of central govemment rather than local expressions ofeffective demand.

2.39 Capital allocation systems vary considerably among countries. In Pakistan, forexample,the official system of capital allocation (described in Box 2) is one that in principle pennits everyproject to be evaluated on both technical and economic grounds, in the lightof the nation's investmentpriorities and the availability of financing. It nevertheless results in project selections being made bypeople far removed from beneficiaries, with little information about projects and less idea ofbeneficiary priorities. The Mexican PRONASOL program, in contrast, is said to institutionalize ahigh degree of beneficiary involvement but tends to bypass the capital budgeting process of sectoral

23

Box 2. Contrasting approaches to capital allocation

ADP: Pak-istan's Annual Developmcnt Plan (ADP) process typifies a highly centralized system that permiiLs everyproject to be cvaluated in the light of thc nation's investment priorities and the availability of financing-but resultsin project selections being madc by people rar removed from the bencficiarics, with little infonnation aboutprojects, and less idca of beneficiary prioritics. Thc ADP process begins with a 'municipality's submission of aproject proposal to the provincial government, wherc it is subjected to technical rcview. If tcchnically approved,it is then included in a larger pool of projects eligible for financing. Financing decisions are made annually, andbegin with an estimate of overall resource availability by the central govemment's Ministry of Finance. Once anoverall division of funds between government and provinces is madc, the provincial govemment makes a tentativematch of resources with projects; it then forwards its recommcndations to the central govemmcnt's annual plancoordinating committee which approves sizc and sectoral allocation of the ovcrall package, and then submits itto a national economic council, presided over by the presidenL This lengthy process succeeds in climninatingtechnically unsound projccts, and matches resources to projects, but incorporates no mechanism for weighing thedegree of local commitment to investment projects.

PRONASOL: Mexico's National Solidarity Program (PRONASOL) represents only the current stage inMexico's gradual process of decentralizing its projectallocation system (starting from a system in which 94 percentof public sector capital investnent decisions were made by sectoral ministries at the federal level). The programis funded from an eannarked share of the national budget (carved out of what used to be sectoral ministry budgets).Allocations are distributed among states by fonnula, with a fixed proportion earmarked for allocation by mayors.Allocations of the municipal share among a state's municipalities are based in part on political considerations. Butwithin a given recipient municipio, the allocation of funds among projects draws on-a well developed system ofnegotiation between the mayor and community groups, in which PRONASOL funding is made conditional on thecommunity's willingness to provide counterpart contributions in cash or in kind. While mayors have the latitudeto vary the tems of each project agreement, the matching requirement is universal. While PRONASOL, as a grantproram, still cmbodies inter-jurisdictional subsidies, ihe explicit inclusion of mayors in the allocation process,and the progranm's use of counterpart matching contribution as a rationing device are, for Mexico, significantreforms.

ministries at the central and state level. In Turkey, the central govermnent's public works agency, IllerBank, is responsible for both financing and construction of municipal public works. Its investmentchoices, similarly, are said to reflect the technical preferences of its engineers, and the politicalinterests of the central government.

III. DIRECTIONS FOR REFORM

3.1 The basis for making recommendations on the reform of intergovernmental relations isnot strong. While the academic literature is useful conceptually, it is notspecifically prescriptive. Theapproach of the developed countries-which might be presumed to demonstrate successful mod-els-provides some common sense of direction, but is still one of considerable variety and frequentexperimentation. The past experience from developing countries appears largely to be a source ofnegative lessons. The most relevant body of evidence is emerging from the countries that are nowin the process of decentralization. Butthese reforms are still in theirinitial stages. There areno before-and-after cases of LDC decentralization on which to base normative advice. The conclusionsprcsented in this chapter are therefore subject to these limitations. They convey more confidence inwhat to avoid than what to pursue, but serve to provide an overall sense of the direction for reform.

3.2 The evidence, such as it is, suggests that there are three elements to reform in thestructure of municipal service delivery: (a) the clarification of functional responsibilities betweenlevels of government; (b) the authorization of revenue sources corresponding to functional respon-sibilities, and (c) the institution of a system of accountability that encompasses both reguiation bycentral govermnent, and incentives for responsiveness to local constituents.

Linking Services to Levels of Government

3.3 Clarity in the division of functional responsibilities between levels of government is anessential condition of any reform in the structure of urban service delivery. Judging from theinstitutional development literature and the anecdotal evidence from developing countries, a clearlinkage between a unit of government and a specific service seems to be critical, in enablingconstituents to hold local governments accountable for specific functions, and in eliminating the softbudget constraint that encourages mayors to act as a lobbyists before the central govemmentministries, rather than as individuals ultimately responsible for the perfornance of specific services.

3.4 Where this clarity does not exist, it would appear to be the first priority for reform, anessential precondition for progress on the other two elements of reform. There can be no correspon-dence between revenue and expenditure assignments unless expenditure responsibilities are known.And any attempt to improve accountability through regulation and constituent access is underminedif the function for which the entity is responsible is undefined.

3.5 Clarity is not achieved by a mere act of legislation or a constitutional demarcation offunctions between tiers of government. The developing world is full of such documents, and they areroutinely ignored or violated. Clarity, above all, requires the central government to refrain from adhoc interventions in responsibilities that have been nominally assigned to local government-toobserve the hard budget constraint with respect to local functions, no matter how disagreeable theoutcomes. And this, in tum, requires a structure of subnational governent that renders such ad hocin